[PATCH v2] RDMA/rtrs: client: Fix clt_path::max_pages_per_mr calculation

[PATCH v2] RDMA/rtrs: client: Fix clt_path::max_pages_per_mr calculation

[Honggang LI]

If device max_mr_size bits in the range [mr_page_shift+31:mr_page_shift]
are zero, the `min3` function will set clt_path::max_pages_per_mr to
zero.

`alloc_path_reqs` will pass zero, which is invalid, as the third parameter
to `ib_alloc_mr`.

v1 -> v2:
Correct the commit message and set max_pages_per_mr to U32_MAX
as Michael Gur suggested.
Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality")
Signed-off-by: Honggang LI <honggangli@163.com>
---
 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
index 71387811b281..e477d2c0ff35 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-clt.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
@@ -1464,6 +1464,8 @@ static void query_fast_reg_mode(struct rtrs_clt_path *clt_path)
 	mr_page_shift      = max(12, ffs(ib_dev->attrs.page_size_cap) - 1);
 	max_pages_per_mr   = ib_dev->attrs.max_mr_size;
 	do_div(max_pages_per_mr, (1ull << mr_page_shift));
+	if ((u32)max_pages_per_mr == 0)
+		max_pages_per_mr = U32_MAX;
 	clt_path->max_pages_per_mr =
 		min3(clt_path->max_pages_per_mr, (u32)max_pages_per_mr,
 		     ib_dev->attrs.max_fast_reg_page_list_len);
-- 
2.52.0

Re: [PATCH v2] RDMA/rtrs: client: Fix clt_path::max_pages_per_mr calculation

[Michael Gur]

On 12/24/2025 11:50 AM, Honggang LI wrote:
> If device max_mr_size bits in the range [mr_page_shift+31:mr_page_shift]
> are zero, the `min3` function will set clt_path::max_pages_per_mr to
> zero.
>
> `alloc_path_reqs` will pass zero, which is invalid, as the third parameter
> to `ib_alloc_mr`.
>
> v1 -> v2:
> Correct the commit message and set max_pages_per_mr to U32_MAX
> as Michael Gur suggested.
> Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality")
> Signed-off-by: Honggang LI <honggangli@163.com>
> ---
>   drivers/infiniband/ulp/rtrs/rtrs-clt.c | 2 ++
>   1 file changed, 2 insertions(+)
>
> diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> index 71387811b281..e477d2c0ff35 100644
> --- a/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> @@ -1464,6 +1464,8 @@ static void query_fast_reg_mode(struct rtrs_clt_path *clt_path)
>   	mr_page_shift      = max(12, ffs(ib_dev->attrs.page_size_cap) - 1);
>   	max_pages_per_mr   = ib_dev->attrs.max_mr_size;
>   	do_div(max_pages_per_mr, (1ull << mr_page_shift));
> +	if ((u32)max_pages_per_mr == 0)
> +		max_pages_per_mr = U32_MAX;
>   	clt_path->max_pages_per_mr =
>   		min3(clt_path->max_pages_per_mr, (u32)max_pages_per_mr,
>   		     ib_dev->attrs.max_fast_reg_page_list_len);

Checking after casting means you'll also assign U32_MAX when it's zero 
before the casting.
For that to happen in this case we'd need to have a device with 
max_mr_size smaller than it's lowest supported page_size.
I think it's safe to assume these are not valid device attributes, so 
we're safe here.

Reviewed-by: Michael Gur <michaelgur@nvidia.com>

Re: [PATCH v2] RDMA/rtrs: client: Fix clt_path::max_pages_per_mr calculation

[Leon Romanovsky]

On Wed, Dec 24, 2025 at 05:50:30PM +0800, Honggang LI wrote:
> If device max_mr_size bits in the range [mr_page_shift+31:mr_page_shift]
> are zero, the `min3` function will set clt_path::max_pages_per_mr to
> zero.
> 
> `alloc_path_reqs` will pass zero, which is invalid, as the third parameter
> to `ib_alloc_mr`.
> 
> v1 -> v2:
> Correct the commit message and set max_pages_per_mr to U32_MAX
> as Michael Gur suggested.

Please put changelog under --- marker, below Signed-off-by tags.

> Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality")
> Signed-off-by: Honggang LI <honggangli@163.com>
> ---
>  drivers/infiniband/ulp/rtrs/rtrs-clt.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> index 71387811b281..e477d2c0ff35 100644
> --- a/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> @@ -1464,6 +1464,8 @@ static void query_fast_reg_mode(struct rtrs_clt_path *clt_path)
>  	mr_page_shift      = max(12, ffs(ib_dev->attrs.page_size_cap) - 1);
>  	max_pages_per_mr   = ib_dev->attrs.max_mr_size;
>  	do_div(max_pages_per_mr, (1ull << mr_page_shift));
> +	if ((u32)max_pages_per_mr == 0)
> +		max_pages_per_mr = U32_MAX;

It is "max_pages_per_mr = min_not_zero((u32)max_pages_per_mr, U32_MAX)"

Thanks

>  	clt_path->max_pages_per_mr =
>  		min3(clt_path->max_pages_per_mr, (u32)max_pages_per_mr,
>  		     ib_dev->attrs.max_fast_reg_page_list_len);
> -- 
> 2.52.0
> 

Re: [PATCH v2] RDMA/rtrs: client: Fix clt_path::max_pages_per_mr calculation

[Honggang LI]

On Thu, Dec 25, 2025 at 02:33:32PM +0200, Leon Romanovsky wrote:
> Please put changelog under --- marker, below Signed-off-by tags.

OK.
 
> >  	do_div(max_pages_per_mr, (1ull << mr_page_shift));
> > +	if ((u32)max_pages_per_mr == 0)
> > +		max_pages_per_mr = U32_MAX;
> 
> It is "max_pages_per_mr = min_not_zero((u32)max_pages_per_mr, U32_MAX)"
> 

Will replace it with `min_not_zero` in v3.

Thanks